Machinist

[Lawrence London <lfljvenaura@gmail.com>]

http://www.reddit.com/r/AskEngineers/comments/23xx6w/how_were_ultraprecise_gears_and_other_mechanical/

[–]magicdragonfly 19 points 3 hours ago

You can make something perfectly flat (for all practical purposes*) by making three of them at the same time.

Take three granite stones, that you've cut to be "mostly flat". Sprinkle rouge on one, take another, and rub the two for a good long time until the two have ground each other down such that the rouge appears fairly evenly distributed. Be sure to rub them in different directions, to distribute the rubbing evenly.

Now, take stones 1 and 3 and do the same thing. You'll notice the rouge rubbing off in certain spots which correspond with the high spots on stone 3, but over time, they'll get knocked down. Then, take stones 2 and 3 and do the same thing. Then go back to 1 and 2, and just keep rotating between stones. Eventually, you'll see that the rouge is very evenly distributed across all three stones, and they've become astonishingly flat and perhaps even taken on a polish. Done properly (this is just the 2 second explanation; search the web for a more in-depth explanation) they'll be flat to within about .0001". For comparison, a sheet of paper (such that you put in a xerox machine) is about .004" thick.

Once you have a flat reference surface, you can make tools that are very precise for making other things. It's also the basis by which you can grind and polish things like gears so they too are flat.

*"Flat" is a tricky thing to attain, because you can always ask "how flat?" and then you discover that with a precise enough measuring tool, what you thought was "flat" was indeed anything but. So you need to temper this with a sense of tolerance -- that is, "what is flat enough for what I'm trying to achieve?" If you're working with wood (which moves around a lot) then "flat enough" could be +/- .005". If you're working with extruded metal, "flat" could be +/- .030" (about 1/32") if you've got a good extrusion.

Also, "flat" and "smooth" are two different things. You can have something that's "flat" be pretty rough, actually, as long as the elevation of each and every high point is exactly the same, and the low points are reasonably few in number that what you're measuring can easily span them.